Intel Labs

It was incredible to see so many fascinating projects being presented at #NeurIPS2024!

Take a look under the hood of CLIP-like vision-language models with CLIP-InterpreT. As one of the most popular types of foundational models, it’s critical we’re able to understand the inner workings of CLIP-like models for tasks such as video retrieval, image generation, and visual navigation. Learn more about the role CLIP-InterpreT plays in our commitment to transparent and responsible AI: https://intel.ly/3DySyYY #VisionLanguageModels #ArtificialIntelligence #ResponsibleAI

Large vision-language models (LVLMs) demonstrate impressive capabilities, but understanding their underlying mechanisms remains a challenging task. As part of our commitment to responsible AI, we collaborated with Microsoft Research Asia to develop LVLM-Interpret, an interactive tool designed to provide detailed visualizations of the inner workings of LVLMs. Learn more here: https://intel.ly/3VHuZ6u #LargeVisionLanguageModels #ArtificialIntelligence #ResponsibleAI

Be at ease when your data is at rest. And in transit. And in use.   Did you know Fully Homomorphic Encryption (FHE) enables data to be processed at each stage in the data life cycle, without it ever being decrypted? Learn about our FHE research here. https://lnkd.in/eEJYjy78   #EncryptedComputing #StartsWithIntel #TechOfTomorrow

As organizations increasingly rely on machine learning for critical decisions, ensuring the integrity and transparency of these systems is vital. A multi-layered framework, developed as an initial proof of concept by our researchers, shows promise in achieving end-to-end AI pipeline security with acceptable performance overhead, paving the way for practical use in production environments. Read the technical deep dive from Marcin Spoczynski, Marcela Melara, and Sebastian Szyller: https://intel.ly/49xqPE1 #MachineLearning #ArtificialIntelligence #Cybersecurity

👉The first foundation model for inductive reasoning that can zero-shot answer logical queries on any knowledge graph. ✅ 👉A novel diffusion-based framework for articulated 3D asset generation. ✅ 👉An encoder-decoder transformer model designed specifically for translating between programming languages and their HPC extensions. ✅ This is just a small sample of the amazing work being showcased at #NeurIPS2024. Check out the blog for more details on the 30+ papers being presented by Intel researchers at this year's event. https://intel.ly/3OOThI0 #NeurIPS2024 #ArtificialIntelligence #NeuralProcessing

To bolster American technology leadership, the US Department of Commerce and Semiconductor Research Corporation (SRC) Manufacturing Consortium Corporation have announced a $285 million award toward chip manufacturing and a new institute, SMART USA (Semiconductor Manufacturing and Advanced Research with Twins USA). SMART USA will focus on efforts to develop, validate, and use digital twins to improve domestic semiconductor design, manufacturing, advanced packaging, assembly, and test processes. Learn more about SMART USA here: https://intel.ly/3OFl6m6 #Manufacturing #Technology #Research

Enhance your AI knowledge at this year's Neural Information Processing Systems (NeurIPS) conference! Don't miss this opportunity to learn directly from Intel Labs researchers about accelerated materials discovery, collaborative AI-driven science, responsible design principles of generative models, and the research achievements of women in machine learning. Check out the blog to learn more about the workshops, demos, and other networking opportunities that will be available. https://intel.ly/49prGXh #AI #NeurIPS #Networking

With the complexities of performance, power, thermals, and more, AI alone isn't ready for chip design. But, it can be a game changer when combined with other innovative design methods. Learn why a hybrid approach could ultimately be the most fruitful path forward in this IEEE Spectrum article from Intel Labs researchers Uday Mallappa, Hesham Mostafa, and Somdeb Majumdar. https://intel.ly/3ZzLvI6 #Manufacturing #Semiconductor #Design

Take a closer look at the industry’s first 1.15 billion neuron neuromorphic system, our path toward more efficient and scalable AI. Codenamed Hala Point, the system uses brain-inspired computing principles to more efficiently solve problems across science, logistics, and large language models. Learn more: https://intel.ly/41dFPVq #ArtificialIntelligence #StartsWithIntel #TechOfTomorrow